Re: Running away from SQL-Ledger

[..hidden..]

> Thanks Tim.

Your welcome! I recall last year when we upgarded to Centos 5.3 and SL 2.6
we were neck deep...the Centos/Perl distribution by default compiles with
usethread.

> We found the info you kindly pointed us to before asking the question.
> In some cases, the error goes away, although it is not related to
> threads -- threads are not in use in our setup. There are reports from
> users that followed the advice, and the problem persisted.

Would be interested in how these users implemented the fix.

> For a concise counter example, see
>
> http://osdir.com/ml/debian-bugs-dist/2010-02/msg07381.html

I for one know for sure the non-usethread Perl is very stable with SL and
we have implemented it for almost a year. I believe our slon deamon also
uses the same local version of non-usethread Perl. No problems reported.

> Note that the developers report that the bug has always been there, yet
> Perl now detects the problem and complains loudly.

It is not surprising that one version of compiler/interpreter would give
warnings to a condition but spilt out errors from a new release. I think
in the SL case it wasn't just complaining. It failed.

> Likely we could compile an older, additional Perl -- this would be a
> better solution than replacing the system's Perl -- but it turned out to
> be way easier for us to keep a chroot with the old system while needed
> (or until a suitable Perl patch is added to Debian's).

We didn't replace the system's Perl or compiled an older verion. We
compiled the newer version without usethread in local and use it as such
(no kidding, in our production system we ran several version of Perls and
Postgreses to support all the apps. It's confusing but okay.)

Best of luck! Tim